Performance improvement of a 3-D configuration reconstruction algorithm using a passive secondary target has been focused in this study. In earlier studies, a theoretical development of the 3-D configuration reconstruction algorithm was achieved and it was implemented by a computer program on a system consisting of an optical bench and a digital imaging system. The passive secondary target used was a circle with two internal spots. In order to use this reconstruction algorithm in autonomous systems, an automatic target recognition algorithm has been developed in this study. Starting from a pre-captured and stored 8-bit gray-level image, the algorithm automatically detects the elliptical image of a circular target and determines its contour in the scene. It was shown that the algorithm can also be used for partially captured elliptical images. Another improvement achieved in this study is the determination of internal camera parameters of the vision system.

Inhaltsangabe:Introduction: Reconstruction of real-world scenes from a set of multiple images is a topic in Computer Vision and 3D Computer Graphics with many interesting applications. There is a relation to Augmented and Mixed Reality (AR/MR), Computer-Supported Collaborative Work (CSCW), Computer-Aided industrial/architectural Design (CAD), modeling of the real-world (e.g. computer games, scenes/effects in movies), entertainment (e.g. 3D TV/Video) and recognition/analyzing of real-world characteristics by computer systems and robots. There exists a powerful algorithm theory for shape reconstruction from arbitrary viewpoints, called shape from photo-consistency. However, it is computationally expensive and hence can not be used with applications in the field of 3D video or CSCW as well as interactive 3D model creation. Attempts have been made to achieve real-time framerates using PC cluster systems. While these provide enough performance they are also expensive and less flexible. Approaches that use GPU hardware-acceleration on single workstations achieve interactive framerates for novel-view synthesis, but do not provide an explicit volumetric representation of the whole scene. The proposed approach shows the efforts in developing a GPU hardware-accelerated framework for obtaining the volumetric photo hull of a dynamic 3D scene as seen from multiple calibrated cameras. High performance is achieved by employing a shape from silhouette technique in advance to obtain a tight initial volume for shape from photo-consistency. Also several speed-up techniques are presented to increase efficiency. Since the entire processing is done on a single PC, the framework can be applied to mobile setups, enabling a wide range of further applications. The approach is explained using programmable vertex and fragment processors and compared to highly optimized CPU implementations. It is shown that the new approach can outperform the latter by more than one magnitude. The thesis is organized as follows: Chapter 1 contains an introduction, giving an overview with classification of related techniques, statement of the main problem, novelty of the proposed approach and its fields of application. Chapter 2 surveys related work in the area of dynamic scene reconstruction by shape from silhouette and shape from photo-consistency. The focus lies on high performance reconstruction and hardware-acceleration. Chapter 3 introduces the theoretical basis for the proposed [...]

"Deep Learning-Based Single View 3D Reconstruction" is a research area that focuses on creating three-dimensional (3D) models of objects from a single two-dimensional (2D) image using deep learning algorithms. This technique has the potential to revolutionize fields such as computer vision, robotics, and virtual reality. The process of reconstructing 3D models from a single image involves several steps, including feature extraction, camera pose estimation, and depth estimation. Deep learning techniques, particularly convolutional neural networks (CNNs), have shown great promise in improving the accuracy and efficiency of these steps. The use of deep learning in single view 3D reconstruction has several advantages over traditional methods, including the ability to learn and generalize complex features and the ability to handle noisy or incomplete data. Additionally, deep learning techniques can be trained on large datasets, allowing for better generalization and improved performance. Henry Smith is not a known author in this field of study, but there are many researchers and experts working on this topic, including David Novotny, Jiri Sedlar, Andrea Vedaldi, and Kostas Daniilidis, among others.

This dissertation, "3D Reconstruction and Camera Calibration From Circular-motion Image Sequences" by Yan, Li, 李燕, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of thesis entitled "3D Reconstruction and Camera Calibration from Circular-Motion Image Sequences" Submitted by Li Yan for the degree of Doctor of Philosophy at The University of Hong Kong in December 2005 This thesis investigates the problem of 3D reconstruction from circular motion image sequences. The problem is normally resolved in two steps: projective reconstruction and then metric reconstruction by self-calibration. A key question considered in this thesis is how to make use of the circular motion information to improve the reconstruction accuracy and reduce the reconstruction ambiguity. The information is previously utilized by identifying the fixed image entities (e.g. the image of the rotation axis, vanishing line of the motion plane, etc). These fixed entities, however, only exist in constant intrinsic parameter sequences. In this thesis, circular motion constraints, which are valid for varying intrinsic parameter (e.g. zooming/refocusing) cameras, are formulated from the movement of camera center and principal plane. Based on the constraints, several novel algorithms are developed for each step of the whole 3D reconstruction procedure. For image sequences with known rotation angles, a circular projective reconstruction algorithm is proposed. We first formulate the circular motion constraints in the Euclidean frame, and then deduce the most general form of reconstruction in a projective frame that satisfies the circular motion constraints. The constraints are gradually enforced during an iterative process, resulting in a circular projective reconstruction. This approach can be used to deal with both cases of constant and varying intrinsic parameters. It is proved that the circular projective reconstruction retrieves metric reconstruction up to a two-parameter ambiguity representing a projective distortion along the rotation axis of the circular motion. Based on the circular projective reconstruction, a hierarchical self-calibration algorithm is proposed to estimate the remaining two parameters. Closed-form expressions of the absolute conic and its image are deduced in terms of the two parameters, which are then determined with zero-skew and unit aspect ratio assumptions. Alternatively, starting from a general (rather than circular) projective reconstruction, a stratified self-calibration algorithm is proposed to upgrade the projective reconstruction directly to a metric one. In this case, the plane at infinity is first identified with (i) the circular motion constraint on camera center and (ii) zero-skew and unit aspect ratio assumptions. With the knowledge of the plane at infinity, the camera calibration matrices can be readily obtained. All the above algorithms assume that the rotation angles are known. In the case if the angles are unknown, we provide two novel rotation angle recovery methods. For constant intrinsic parameter sequences, rotation angles can be recovered by using the fixed image entities. For varying intrinsic parameter sequences, it is shown that the movements of the camera center and principal plane form two concentric circles on the motion plane. By identifying the corresponding conic loci in 3D projective frame, the geometry of circular motion on the motion plane can be recovered. Compared with existing methods, the new method is more flexible in that it allows the intrinsic parameters to vary, and is simpler by avoi

The issue discusses methods to extract 3-dimensional (3D) models from plain images. In particular, the 3D information is obtained from images for which the camera parameters are unknown. The principles underlying such uncalibrated structure-from-motion methods are outlined. First, a short review of 3D acquisition technologies puts such methods in a wider context, and highlights their important advantages. Then, the actual theory behind this line of research is given. The authors have tried to keep the text maximally self-contained, therefore also avoiding to rely on an extensive knowledge of the projective concepts that usually appear in texts about self-calibration 3D methods. Rather, mathematical explanations that are more amenable to intuition are given. The explanation of the theory includes the stratification of reconstructions obtained from image pairs as well as metric reconstruction on the basis of more than 2 images combined with some additional knowledge about the cameras used. Readers who want to obtain more practical information about how to implement such uncalibrated structure-from-motion pipelines may be interested in two more Foundations and Trends issues written by the same authors. Together with this issue they can be read as a single tutorial on the subject.

The fully automated estimation of the 6 degrees of freedom camera motion and the imaged 3D scenario using as the only input the pictures taken by the camera has been a long term aim in the computer vision community. The associated line of research has been known as Structure from Motion (SfM). An intense research effort during the latest decades has produced spectacular advances; the topic has reached a consistent state of maturity and most of its aspects are well known nowadays. 3D vision has immediate applications in many and diverse fields like robotics, videogames and augmented reality; and technological transfer is starting to be a reality. This book describes one of the first systems for sparse point-based 3D reconstruction and egomotion estimation from an image sequence; able to run in real-time at video frame rate and assuming quite weak prior knowledge about camera calibration, motion or scene. Its chapters unify the current perspectives of the robotics and computer vision communities on the 3D vision topic: As usual in robotics sensing, the explicit estimation and propagation of the uncertainty hold a central role in the sequential video processing and is shown to boost the efficiency and performance of the 3D estimation. On the other hand, some of the most relevant topics discussed in SfM by the computer vision scientists are addressed under this probabilistic filtering scheme; namely projective models, spurious rejection, model selection and self-calibration.